動(dòng)力學(xué)數(shù)據(jù)的經(jīng)驗(yàn)建模與比較

化工論文編號(hào):HG109  論文字?jǐn)?shù):17014,頁數(shù):37

目  錄
中文摘要 I
英文摘要 Ⅱ
目  錄 III
1． 緒論 1
 1.1  動(dòng)力學(xué)數(shù)據(jù)處理的應(yīng)用背景與意義 1
 1.2  動(dòng)力學(xué)數(shù)據(jù)處理的發(fā)展和研究現(xiàn)狀 2
 1.3  動(dòng)力學(xué)數(shù)據(jù)處理基本思想 2
 1.4  常用動(dòng)力學(xué)數(shù)據(jù)處理方法 2
2． 熱分析法 6
 2.1  常用熱分析法 6
 2.2  小結(jié) 8
3． 微分法和積分法 9
 3.1  常用微分法和積分法 9
 3.2  小結(jié) 10
4． 人工神經(jīng)網(wǎng)絡(luò) 11
 4.1  神經(jīng)網(wǎng)絡(luò)建模方法思想概述 11
 4.2  神經(jīng)元 17
 4.3  互連模式 18
 4.4  學(xué)習(xí)規(guī)則 19
 4.5  神經(jīng)網(wǎng)絡(luò)的建模方法 21
 4.6  小結(jié) 22
5． 回歸分析 23
 5.1  多元線性回歸（MLR） 23
 5.2  小結(jié) 24
6． 實(shí)例論證和結(jié)果討論 25
 6.1  實(shí)例論證 25
 6.2  結(jié)果討論 28
7． 總結(jié)與展望 30
 7.1  總結(jié) 30
 7.2  展望 30
致謝 31
參考文獻(xiàn) 32

摘要：動(dòng)力學(xué)參數(shù)求解是動(dòng)力學(xué)數(shù)據(jù)處理問題的主要內(nèi)容和關(guān)鍵技術(shù)。本文介紹了熱分析法，微積分法，回歸分析，人工神經(jīng)網(wǎng)絡(luò)方法通過求解動(dòng)力學(xué)參數(shù)，最終達(dá)成動(dòng)力學(xué)模型的經(jīng)驗(yàn)建模和比較。最后選用人工神經(jīng)網(wǎng)絡(luò)建模方法和回歸分析建模方法，以醋酸乙酯的皂化反應(yīng)的動(dòng)力學(xué)數(shù)據(jù)為試驗(yàn)對(duì)象，用MATLAB編程計(jì)算結(jié)果。通過對(duì)實(shí)驗(yàn)數(shù)據(jù)結(jié)果的對(duì)比分析，可以得知人工神經(jīng)網(wǎng)絡(luò)在減少結(jié)果誤差，提高分析結(jié)果準(zhǔn)確度方面顯示出了很好的優(yōu)勢(shì)。

關(guān)鍵詞：動(dòng)力學(xué)數(shù)據(jù)；參數(shù)估計(jì)；人工神經(jīng)網(wǎng)絡(luò)；回歸分析；經(jīng)驗(yàn)建模

Experiential Modeling and Comparison of Kinetics Data

Abstract：Parameters Solution of Kinetic is the main contents and key technologies of the problem of Kinetics data process. This paper introduces thermal Analysis, infinitesimal calculus method, regression analysis as well as artificial neural network and then achieves the objective of Kinetics data empirical modeling and comparison by solving parameter. Finally, with artificial neural network and regression analysis selected and Kinetics data of Saponification Reaction of ethyl acetate, results were calculated based on MATLAB programming. Through comparison analysis of the results of experimental data, it is believed that artificial neural network has shown the remarkable superiority at the aspects of reducing result error and improving the accuracy of the results of the analysis.

Keywords: Kinetic data;Parameter estimation;Artificial neural networks; Regression analysis; The experience of modeling